If that's a public site or you know a similar public site and it can be used as a visitor then share a link so we could troubleshoot the problem more reliably.
You can try this regex:
[A-Z]+(?<=ALL)
That worked for me great when I tested with a dummy site.
I also tested with this:
(?:^|\W)ALL(?:$|\W)
That worked but for some reason it always goes to the one word above the "ALL" word. So in this case "*Messages"
Even if I changed the OCR to find "ALLY" then the mouse hovered over "CLIENT VIEW". Same thing happened if I didn't use regex but just had spaces surrounding all. Like so:
I don't know if its some weird bug in PAD. My version is 2.52.6225009.
If you were to use this then you would want to add Offset Y: to something like 15 so it goes that many pixels down. Not the most reliable solution I know.
With Move mouse to image the image has to be exactly as you have captured it. So you have to take the ALL (and some space around it so it doesn't match lets say "MALL" or "ALLY" if there could be those capitalized words on the screen with same font and font size.
Like so:
But if you capture it like this for example:
Then if the *Messages is not there sometimes or CLIEN VIEW is different then it will fail. You might have already known this and the image ocr is just being unstable.
Is there no way to use browser actions? If not have you tried UI automation actions? Sometimes those work even on websites. Even if it only gives you the overall Pane -element and not the specific items you would choose from the drop-down. Open the dropdown first and then use "Set drop-down list value in window" -action and select options by name and use the "ALL"
